pineapple-neural 0.0.5

Simplified processing for image-based cell profiling with pineapple
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158

// Copyright (c) 2025, Tom Ouellette
// Licensed under the BSD 3-Clause License

use candle_core::Device;
use candle_core::{Module, Result, Tensor};

use pineapple_core::im::PineappleImage;

use crate::models::DinoVisionTransformer;
use crate::models::StandardVisionTransformer;

use crate::load::{
    load_dinobloom_vit_base, load_dinov2_vit_base, load_dinov2_vit_small, load_scdino_vit_small,
    load_subcell_vit_base,
};

use crate::preprocess::{preprocess_imagenet, preprocess_subcell};

pub enum Models {
    DinoVitSmall(DinoVisionTransformer),
    DinoVitBase(DinoVisionTransformer),
    DinobloomVitBase(DinoVisionTransformer),
    ScdinoVitSmall(StandardVisionTransformer),
    SubcellVitSmall(StandardVisionTransformer),
}

impl Models {
    pub fn load(model_name: &str, device: &Device, verbose: bool) -> Self {
        match model_name {
            "dino_vit_small" => {
                let model = load_dinov2_vit_small(device, verbose).unwrap();
                Models::DinoVitSmall(model)
            }
            "dino_vit_base" => {
                let model = load_dinov2_vit_base(device, verbose).unwrap();
                Models::DinoVitBase(model)
            }
            "dinobloom_vit_base" => {
                let model = load_dinobloom_vit_base(device, verbose).unwrap();
                Models::DinobloomVitBase(model)
            }
            "scdino_vit_small" => {
                let model = load_scdino_vit_small(device, verbose).unwrap();
                Models::ScdinoVitSmall(model)
            }
            "subcell_vit_base" => {
                let model = load_subcell_vit_base(device, verbose).unwrap();
                Models::SubcellVitSmall(model)
            }
            _ => {
                eprintln!("[pineapple::nn::models] Model name not found.");
                std::process::exit(1);
            }
        }
    }

    pub fn preprocess(&self, image: &PineappleImage, device: &Device) -> Result<Tensor> {
        match self {
            Models::DinoVitSmall(_) => preprocess_imagenet(image, device),
            Models::DinoVitBase(_) => preprocess_imagenet(image, device),
            Models::DinobloomVitBase(_) => preprocess_imagenet(image, device),
            Models::ScdinoVitSmall(_) => preprocess_imagenet(image, device),
            Models::SubcellVitSmall(_) => preprocess_subcell(image, device),
        }
    }

    pub fn forward(&self, input: &Tensor) -> Result<Tensor> {
        let input = input.unsqueeze(0).unwrap();
        match self {
            Models::DinoVitSmall(model) => model.forward(&input),
            Models::DinoVitBase(model) => model.forward(&input),
            Models::DinobloomVitBase(model) => model.forward(&input),
            Models::ScdinoVitSmall(model) => model.forward(&input),
            Models::SubcellVitSmall(model) => model.forward(&input),
        }
    }
}

#[cfg(test)]
mod test {
    use super::*;
    use pineapple_core::im::PineappleImage;

    fn load_rgb() -> PineappleImage {
        PineappleImage::open("../data/tests/test_rgb.png").unwrap()
    }

    fn load_grayscale() -> PineappleImage {
        PineappleImage::open("../data/tests/test_grayscale.png").unwrap()
    }

    fn test_model(name: &str, color: &str, n_embed: usize) {
        let image = if color == "rgb" {
            load_rgb()
        } else {
            load_grayscale()
        };

        let model = Models::load(name, &Device::Cpu, true);
        let image = model.preprocess(&image, &Device::Cpu).unwrap();
        let logits = model.forward(&image).unwrap();

        let (n_row, n_columns) = logits.shape().dims2().unwrap();

        assert_eq!(n_row, 1);
        assert_eq!(n_columns, n_embed);
    }

    #[test]
    fn test_dinov2_small_rgb() {
        test_model("dino_vit_small", "rgb", 384);
    }

    #[test]
    fn test_dinov2_small_grayscale() {
        test_model("dino_vit_small", "grayscale", 384);
    }

    #[test]
    fn test_dinov2_base_rgb() {
        test_model("dino_vit_base", "rgb", 768);
    }

    #[test]
    fn test_dinov2_base_grayscale() {
        test_model("dino_vit_base", "grayscale", 768);
    }

    #[test]
    fn test_dinobloom_rgb() {
        test_model("dinobloom_vit_base", "rgb", 768);
    }

    #[test]
    fn test_dinobloom_grayscale() {
        test_model("dinobloom_vit_base", "grayscale", 768);
    }

    #[test]
    fn test_subcell_rgb() {
        test_model("subcell_vit_base", "rgb", 768);
    }

    #[test]
    fn test_subcell_grayscale() {
        test_model("subcell_vit_base", "grayscale", 768);
    }

    #[test]
    fn test_scdino_rgb() {
        test_model("scdino_vit_small", "rgb", 384);
    }

    #[test]
    fn test_scdino_grayscale() {
        test_model("scdino_vit_small", "grayscale", 384);
    }
}